Occurrences and Characterization of Antibiotic-Resistant Bacteria and Genetic Determinants of Hospital Wastewater in a Tropical Country

Abstract

Wastewater discharged from clinical isolation and general wards at two hospitals in Singapore was examined to determine the emerging trends of antibiotic resistance (AR). We quantified the concentrations of 12 antibiotic compounds by analysis using liquid chromatography-tandem mass spectrometry (LC-MS/MS), antibiotic-resistant bacteria (ARB), the class 1 integrase gene (intI1), and 16 antibiotic resistance genes (ARGs) that confer resistance to 10 different clinically relevant antibiotics. A subset of 119 antibiotic-resistant isolates were phylogenetically classified and tested for the presence of ARGs encoding resistance to β-lactam antibiotics (bla_NDM, bla_KPC, bla_SHV, bla_CTX-M), amikacin [aac(6′)-Ib], co-trimoxazole (sul1, sul2, dfrA), ciprofloxacin (qnrA, qnrB), and the intI1 gene. Among these resistant isolates, 80.7% were detected with intI1 and 66.4% were found to carry at least 1 of the tested ARGs. Among 3 sampled locations, the clinical isolation ward had the highest concentrations of ARB and the highest levels of ARGs linked to resistance to β-lactam (bla_KPC), co-trimoxazole (sul1, sul2, dfrA), amikacin [aac(6′)-Ib], ciprofloxacin (qnrA), and intI1. We found strong positive correlations (P < 0.05) between concentrations of bacteria resistant to meropenem, ceftazidime, amikacin, co-trimoxazole, and ciprofloxacin and abundances of bla_KPC, aac(6′)-Ib, sul1, sul2, dfrA, qnrA, and intI1 genes.

INTRODUCTION

The emergence of antibiotic resistance (AR) has led to heightened global concern due to the reduced efficacy of currently available antibiotics in the treatment of bacterial infections (1, 2). Gram-negative pathogens such as Escherichia coli and Klebsiella pneumoniae have shown increased resistance to a variety of antibiotics typically used in the treatment of infections and diseases caused by these bacteria (3). Hospital wastewater is considered a hot spot for AR as a consequence of receiving a cocktail of antibiotic compounds, disinfectants, and inputs of bacterial sheddings and metabolized drugs from patient excrement, which potentially contain multidrug-resistant (MDR) pathogens (4–6). As such, hospital wastewaters provide an environment for the exchange of antibiotic resistance genes (ARGs) between clinical pathogens and other environmental bacteria in recipient sewers, which could result in broader epidemiological consequences extending beyond the hospital setting.

Integrons enable the recombination and expression of mobile gene cassette arrays which contain ARGs and are used as indicators to gather information on trends of AR development and dissemination in bacterial communities (7, 8). Gram-negative bacteria bearing multiple bla genes (e.g., bla_NDM, bla_KPC, bla_CTX-M, and bla_SHV) are a problem and are increasingly found in hospital wastewaters (4, 9), together with other common and related drug-resistant ARGs such as qnr (fluoroquinolone), erm (macrolide), sul (sulfonamides), and tet (tetracycline). High densities of MDR bacteria in hospital wastewaters could facilitate the propagation and dissemination of ARGs by horizontal gene transfer via plasmids, transposons, and integrons (8, 10–12). In a preventive effort aimed at reducing the spread of antimicrobial resistance and formulating intervention strategies, clinical institutions have taken steps to conduct surveillance studies on antibiotic prescriptions in hospitals through longitudinal assessments of antibiotic prescription patterns (13). More specifically, linking information from monitoring studies of the fate of antibiotics to data on the occurrence of the type and abundances of antibiotic-resistant bacteria (ARB) and antibiotic resistance genes (ARGs) will provide a direct and concise measure of dissemination patterns of antibiotic resistance.

The objective of this study was to quantify antibiotic residues (AB) using liquid chromatography-tandem mass spectrometry (LC-MS/MS), antibiotic-resistant bacteria (ARB) using a culture-based method, and antibiotic resistance genes (ARGs) using quantitative PCR (qPCR) in hospital wastewaters in Singapore. This improves our current knowledge of the potential impacts of hospital wastewater on antibiotic resistance in the environment under tropical conditions. The correlation between these antibiotic resistance factors and the class 1 integron was analyzed to evaluate the potential dissemination of multidrug resistance across the microbial community. The results provide important baseline data and highlight the more prevalent ARB and ARGs found in local hospitals in Asia, as well as levels of antibiotics present.

MATERIALS AND METHODS

Collection of hospital wastewater samples.

Eight clinical wastewater samples were collected from two hospitals in Singapore. Two hospital blocks (i.e., block a of hospital 1 [H1a] and block b of hospital 1 [H1b]) (1,597 beds) were sampled once a week over a period of 2 weeks from a manhole receiving direct sewage from each block. These two blocks were differentiated based on their ward types, with block A (i.e., H1a) consisting of clinical isolation wards and block B (i.e., H1b) consisting of more-general wards. For hospital 2 (i.e., H2; 1,500 beds), samples were collected once a week over a period of 1 month from the main manhole discharging mixed wastewater from the entire hospital. All samples were collected on weekdays in the morning between 9:30 a.m. and 11:30 a.m. during September 2014 at ambient temperatures ranging from 24°C to 31°C. A volume of 1 liter of wastewater was collected from each site using sterile plastic bottles, preserved on ice, and transported to the laboratory (within 2 h) for chemical and microbiological analyses. The 1-liter sample was divided into two 0.5-liter samples which were reserved for DNA extraction and chemical analysis.

DNA extraction.

A volume of 0.5 liter of hospital wastewater was filtered through a 0.45-μm-pore-size cellulose nitrate membrane (Sartorius Stedim, Gottingen, Germany). Total DNA was extracted from the biomass using a PowerWater DNA isolation kit (Mo Bio Laboratories, Inc., Carlsbad, CA) according to the manufacturer's instructions. DNA concentration and purity were measured using a NanoDrop spectrophotometer (NanoDrop 2000; Thermo Fisher Scientific, Wilmington, DE, USA).

Quantification of antibiotics.

The 0.5 liter of hospital wastewater samples was processed immediately upon arrival in the laboratory by filtration through a 0.1-μm-pore-size Millipore glass fiber filter (Millipore, Bedford, MA, USA) and adjusted to a pH of 2.4. Extraction of antibiotics was performed using two solid-phase extraction (SPE) cartridges in series: first with a 500-mg LC-NH₂ (Supelco, Sigma-Aldrich, MO, USA) SPE anionic-exchange cartridge to remove part of the impurities and then with a polymeric 500-mg Oasis HLB (Waters, Milford, CT) to retain target compounds. Final compounds were eluted with 9 ml of methanol in amber glass vials, dried under a stream of nitrogen, and reconstituted with 0.9 ml methanol. An LC-electrospray ionization (ESI)-MS/MS analysis performed on an Agilent 1290 Infinity LC coupled to a 6490 Triple Quad MS/MS system (Agilent Technologies Inc. Santa Clara, CA) was used to detect concentrations of 12 antibiotics in the hospital wastewater samples: meropenem (MEM), lincomycin, sulfamethazine, trimethoprim, tetracycline, ciprofloxacin (CIP), sulfamethoxazole, chloramphenicol (CHL), azithromycin, clindamycin (CLI), erythromycin (ERI), and clarithromycin. To identify these antibiotic targets, the retention times and the area ratios of the qualitative and quantitative ions in each sample were compared with those of the calibration standards. The results were accepted when (i) the retention time was within ±15 s of the average retention time in the calibration standards; (ii) the uncertainty of two product ions was within ±20% of that in the standards; and (iii) the signal/noise (S/N) ratio of the quantitative ions was higher than 3. The quantitation process followed USEPA method 1694 and was performed in duplicate.

The SPE recoveries (see Table S1 in the supplemental material) were tested by spiking four replicates of reagent water with antibiotics and processing the spiked waters through the entire analytical method. The overall recoveries, taking into consideration both the SPE and matrix recoveries, were tested by spiking four different sewage samples with antibiotics and analyzing the spiked samples through the entire analytical method. Antibiotics (in methanol) were spiked at 1 μg/liter. The procedural recoveries and overall recoveries and their relative standard deviations are listed in Table S1. Samples were analyzed in duplicate; one procedural blank consisting of reagent water was processed together with each batch of sample to demonstrate that the analytical system and glassware were free of contamination.

Quantification of ARB.

To determine the concentration of bacteria resistant to each of the 10 antibiotics tested in the study, hospital wastewater samples were serially diluted and cultured on LB agar (BD Diagnostics, Sparks, MD, USA) supplemented with each of the 10 antibiotics. Antibiotics (Sigma-Aldrich, USA) were used in the screening process at the following concentrations: amikacin (AMK) at 17 μg/ml; meropenem (MEM) at 4 μg/ml; ceftazidime (CAZ) at 8 μg/ml; clindamycin (CLI) at 0.5 μg/ml; erythromycin (ERY) at 0.5 μg/ml; ciprofloxacin (CIP) at 2 μg/ml; co-trimoxazole (SXT; trimethoprim at 3 μg/ml and sulfamethoxazole at 57 μg/ml); tetracycline (TET) at 4 μg/ml; vancomycin (VAN) at 4 μg/ml; and chloramphenicol (CHL) at 8 μg/ml. All of the antibiotic concentrations used for screening were higher than the susceptible breakpoints published in the 2013 Clinical and Laboratory Standards Institute (CLSI) Performance Standards for Antimicrobial Susceptibility Testing report, except for the concentrations of erythromycin and clindamycin, which were used at their breakpoints. Raw water samples were serially diluted in 1× phosphate-buffered saline (PBS) (Vivantis Technologies, Malaysia), and 100-μl volumes of samples were spread-plated in triplicate and incubated at 35°C for 24 h and subsequently at room temperature the next day for an additional 24 h. Dilution-containing colonies within the range of 30 to 100 cells were counted in order to minimize the error caused by the too-numerous-to-count problem.

Quantification of ARGs.

Specific primers for real-time PCR were selected from the literature and designed in this study for quantification of the 16S rRNA gene (14, 15), intI1 (16), and 16 other ARGs, including bla_KPC (17), bla_NDM (this study), bla_SHV (18), bla_CTX-M (19), sul1 (this study), sul2 (20), dfrA (21), aac(6′)-Ib (this study), tet(M) (22), tet(O) (23), erm(B) (24), mef(A/E) (25), cfr (26), vanA (this study), qnrA (27), and qnrB (27). These genes confer resistance to the 10 antibiotics used in our phenotypic resistance screening (see Table S2 in the supplemental material). Total genomic DNA directly extracted from hospital wastewaters (see “DNA extraction” section) was used for qPCR assays, and analysis of each sample was performed in triplicate within the same run together with a calibration curve and a no-template control (NTC). The calibration curves were built using a 10-fold dilution series of ARG standards. Each qPCR mixture contained 10 μl SYBR Select master mix (Applied Biosystems, CA, USA), 0.5 μl each of 10 μM concentrations of forward primers and reverse primers, and 20 to 50 ng of DNA template and was made up to a volume of 20 μl with PCR-grade water. The qPCR assays were conducted in a StepOnePlus real-time PCR system (Applied Biosystems, CA, USA) under the following conditions: holding at 50°C for 2 min and 95°C for 3 min; 40 cycles of 95°C for 15 s and annealing temperature for 1 min; and, finally, a melting curve analysis performed at 95°C for 15 s, 60°C for 1 min, and 95°C for 15 s. Only amplicons with a single peak in the melting curve analysis were considered specific amplified products of targeted genes. All qPCR assays were performed according to the manufacturer's specifications. Calculation of the absolute gene copy numbers followed a previously described procedure (28, 29). Relative gene abundances were calculated by normalizing the absolute number of ARG copies to that of 16S rRNA gene copies.

Identification and ARG detection of ARB isolates.

A total of 236 colonies were randomly picked from plates containing 1 of the 10 antibiotics and were subcultured to ensure that pure isolates were obtained. Single colonies were inoculated into liquid LB broth and allowed to grow overnight at 37°C, and 10 μl of culture suspension was subsequently subjected to 16S rRNA gene PCR amplification using the 27F (5′-AGA GTT TGA TYM TGG CTC AG-3′) and 1492R (5′-GGY TAC CTT GTT ACG ACT T-3′) primers. The PCR products were purified using an Expin cleanup kit (GeneAll Biotechnology, Seoul, South Korea) and sent for capillary sequencing at AITbiotech (Singapore). Sequences were manually edited using Bioedit software and queried against the National Center for Biotechnology Information (NCBI) 16S rRNA gene database for archaea and bacteria. The bacterial gene sequences were deposited in GenBank (see below for accession numbers).

In addition, bacterial isolates showing phenotypic resistance to the following antibiotics were screened for the presence of associated ARGs (in parentheses): β-lactam (bla_NDM, bla_KPC, bla_CTX-M, bla_SHV), amikacin [aac(6′)-Ib], co-trimoxazole (sul1, sul2, dfrA), ciprofloxacin (qnrA, qnrB), and integrase 1 (intI1). Conventional PCR assay was used, and the working conditions were as follows: 95°C for 5 min; 35 repeats of a cycle of 95°C for 30 s, an annealing temperature for 30 s, and 72°C for 30 s; and, finally, 72°C for 10 min. The optimal annealing temperature for each primer was tested with a temperature gradient from 55°C to 65°C, as shown in Table S2 in the supplemental material. The PCR products were sequenced and queried against the NCBI nonredundant database using BLAST.

Statistical analysis.

Statistical analysis was performed using a software package (Primer version 6 and Origin version 9.0). For every analysis of AB, ARB, and ARGs, three replicates were performed, and the averages of the results were calculated. Differences in average antibiotic, ARB, and ARG concentrations among samples were statistically verified using analysis of variance (ANOVA) and/or Student's t test. Correlations between ARB concentrations, ARG abundances, and antibiotic concentrations were analyzed using Spearman and Pearson tests. A principal component analysis (PCA) was employed to assess relationships between measured concentrations of antibiotic residues and cultivated ARB and ARGs for each sampled location.

Accession number(s).

The bacterial gene sequences determined in this work were deposited in GenBank with accession numbers KU312692 to KU312927.

RESULTS

Occurrence of antibiotic residues in hospital wastewater.

Trimethoprim, sulfamethoxazole, erythromycin, azithromycin, and clarithromycin were detected in all samples at various concentrations fluctuating between 0.78 and 71.8 μg/liter, 0.94 and 28.36 μg/liter, 0.30 and 17.63 μg/liter, 0.11 and 1.51 μg/liter, and 0.76 and 72.87 μg/liter, respectively, with azithromycin having the lowest concentration among the five antibiotics (Table 1). Ciprofloxacin was detected at higher levels in location H1b, and tetracycline was detected only in location H2 at a low level whereas meropenem, lincomycin, sulfamethazine, chloramphenicol, and clindamycin levels were relatively low or below the method detection limit (MDL). As a general trend, the concentrations of antibiotics in the location H1b samples were slightly higher than those in the wastewater samples from the two other locations.

TABLE 1.

Concentration range of antibiotics detected for hospital wastewater samples^a

Antibiotic	Concentration(s) (μg/liter)			Method detection limit (μg/liter)
	Location H1a	Location H1b	Location H2
Azithromycin	0.18–1.511	0.13–1.16	0.11–0.72	0.0036
Chloramphenicol	<MDL–0.65	<MDL	<MDL	0.0378
Ciprofloxacin	<MDL	8.74–76.44	1.72–4.34	0.1405
Clarithromycin	1.29–20.49	0.76–72.87	0.76–5.11	0.0185
Clindamycin	<MDL	<MDL	<MDL–1.43	0.0559
Erythromycin	1.64–8.09	0.49–13.71	0.3–17.63	0.0332
Lincomycin	<MDL	<MDL	<MDL–0.015	0.0030
Meropenem	<MDL–0.19	<MDL–1.01	<MDL–1.07	0.0303
Sulfamethazine	<MDL	<MDL	<MDL	0.0160
Sulfamethoxazole	19.90–24.43	7.21–28.36	0.94–23.92	0.0445
Tetracycline	<MDL	<MDL	<MDL–0.82	0.1017
Trimethoprim	8.63–13.83	6.61–71.8	0.78–11.87	0.0155

^aH1a, hospital 1, at block A; H1b, hospital 1, at block B; H2, hospital 2, at the mixed manhole (i.e., the main manhole discharging mixed wastewater from the entire hospital); MDL, method detection limit.

Occurrence of ARB in hospital wastewater.

The average concentration of cultivable ARB in the H1a samples (7.64 × 10⁶ CFU/ml) was 6-fold higher than that in the H2 samples (1.30 × 10⁶ CFU/ml) and 10-fold higher than that in the H1b samples (7.57 × 10⁵ CFU/ml). The concentrations of ARB exhibiting resistance to specific classes differed across different hospitals and ward types; however, across all samples, the average concentrations of ARB resistant to amikacin (1.06 × 10⁶ CFU/ml), clindamycin (1.37 × 10⁶ CFU/ml), erythromycin (1.24 × 10⁶ CFU/ml), ciprofloxacin (1.14 × 10⁶ CFU/ml), and tetracycline (1.30 × 10⁶ CFU/ml) were at least 1 order of magnitude higher than those of ARB resistant to meropenem (4.79 × 10⁵ CFU/ml), ceftazidime (8.22 × 10⁵ CFU/ml), vancomycin (9.19 × 10⁵ CFU/ml), chloramphenicol (6.08 × 10⁵ CFU/ml), and co-trimoxazole (2.54 ×10⁵ CFU/ml) (Fig. 1). The concentrations of ARB for each of the 10 antibiotics in sample H1a were consistently higher than those determined in the other samples (ANOVA and Tukey test, P < 0.05).

FIG 1.

Concentration (in CFU per milliliter) of antibiotic-resistant bacteria in hospital wastewaters for the locations of clinical isolation ward H1a (black), general ward H1b (white), and mixed wastewater at hospital H2 (gray).

Occurrence of ARGs in hospital wastewater.

The abundances of all 16 ARGs corresponding to 10 antibiotic classes in the hospital wastewaters were quantified using real-time PCR assay. The average concentrations of ARGs ranged between 10⁴ and 10⁹ gene copies/ml, with the exception of the cfr gene, which was the least prevalent, with an average concentration of 11.0 gene copies/ml (Fig. 2; see also Fig. S1 in the supplemental material). The 16S rRNA gene was the most abundant across all samples, with an average concentration of 1.58 × 10⁸ gene copies/ml (Fig. 2; see also Fig. S1). The integrase intI1 gene, which is used as a proxy for the horizontal gene transfer (HGT) of ARGs, was detected at an average concentration of 2.31 × 10⁶ gene copies/ml.

FIG 2.

Concentration (in numbers of copies per milliliter) of ARGs at H1a (black), H1b (diagonal lines), and H2 (gray). Significant differences (P value < 0.05 [Student's t test]) are indicated by different letters (i.e., a and b, a and bc) at the top of each bar. The absence of significant differences is indicated by identical letters (i.e., a and a, a and ab).

The four ARG targets for β-lactam resistance, including bla_NDM, bla_KPC, bla_CTX-M, and bla_SHV, were detected at average concentrations of 2.29 × 10⁶, 4.08 × 10⁷, 1.25 × 10⁶, and 6.19 × 10⁵ gene copies/ml, respectively (Fig. 2). bla_NDM and bla_KPC were slightly more prevalent than bla_CTX-M and bla_SHV in the hospital wastewaters. The relative abundances of ARGs were calculated by normalizing the gene copy numbers of each ARG to the 16S rRNA gene copy numbers in the same samples (see Fig. S1 in the supplemental material). Among the four bla genes, bla_KPC was the most prevalent gene in H1a, while bla_NDM was the most prevalent in H1b, and no significant differences in bla_CTX-M and bla_SHV levels were found between the isolation ward (H1a) and general ward (H1b) locations.

Identification and ARG detection of ARB isolates.

Colony PCR assays of meropenem- and ceftazidime-resistant isolates (n = 48) indicated the percentages of the total isolates carrying the bla_NDM, bla_KPC, bla_CTX-M, and bla_SHV genes (see Table S3 and S4 in the supplemental material). Most isolates carrying bla_KPC were of the Enterobacteriaceae family (i.e., Enterobacter spp., Pantoea spp., K. pneumoniae, and Kluyvera georgiana), while those harboring bla_NDM belonged to both Enterobacteriaceae and non-Enterobacteriaceae species (i.e., Pseudomonas spp., Acinetobacter spp., and Comamonas testosteroni).

Among the co-trimoxazole-resistant isolates (n = 24), which were tested for sul1, sul2, and drfA ARGs, sul1 was the most predominant gene detected (at 83.3%), while the other two ARGs, sul2 and dfrA, were not found in any isolates. This pattern of detection in isolates reflects the gene prevalence in the hospital wastewater determined by qPCR, in which sul1 was the most dominant co-trimoxazole resistance gene (see Table S5 in the supplemental material). Most of the amikacin-resistant isolates (n = 22) were positive for the aac(6′)-Ib (95.5%) and intI1 (86.4%; see Table S6) genes, while most of the ciprofloxacin-resistant isolates (n = 23) were negative for the qnrA and qnrB genes, suggesting that another mechanism may be implicated in ciprofloxacin resistance for these particular strains (see Table S7).

Correlations between antibiotic residues, ARB, and ARGs.

Both Pearson and Spearman correlation analyses were employed to examine correlations between the concentrations of antibiotic residues, ARB (in CFU per milliliter), and the associated ARGs (in numbers of gene copies per milliliter) for meropenem, ceftazidime, co-trimoxazole, amikacin, and ciprofloxacin antibiotics. We found positive correlations between the concentrations of (i) MEM (r = 0.984, P = 1.06 × 10⁻⁵) and CAZ (r = 0.709, P = 4.91 × 10⁻²) and ARB and bla_KPC genes; (ii) SXT ARB and sul1 (r = 0.937, P = 6.07 × 10⁻⁴), sul2 (r = 0.994, P = 5.03 × 10⁻⁷), and dfrA (r = 0.910, P = 1.68 × 10⁻³) genes; (iii) AMK ARB and the aac(6′)-Ib (r = 0.872, P = 4.8 × 10⁻³) gene; and (iv) CIP ARB and the qnrA (r = 0.977, P = 3.06 × 10⁻⁵) gene (Table 2, Table 3, and Table 4). Among the antibiotics measured in the hospital wastewaters, significant correlations were found only between ciprofloxacin concentrations and qnrB genes (r = 0.994). The intI1 gene, a potential ARG-transferring vector, showed positive correlations with the bla_SHV, sul1, dfrA, and aac(6′)-Ib genes. A PCA plot (see Fig. S2 in the supplemental material) indicated that H1a was one of locations with the highest sources of antibiotic resistance in terms of concentrations of ARB, intI1, and 7 of the 16 ARGs tested [the bla_KPC, bla_CTX-M, sul1, sul2, dfrA, aac(6′)-Ib, and qnrA genes]. Wastewaters from this location (H1a), however, did not have the highest measured levels of ciprofloxacin, sulfamethoxazole, and trimethoprim antibiotics, which in this case were higher at H1b (Table 1). Other ARGs such as bla_SHV, bla_NDM, and qnrB were more abundant at other locations (see Fig. S1 in the supplemental material).

TABLE 2.

Correlation between bla_NDM, bla_KPC, bla_CTX-M, and bla_SHV genes, meropenem and ceftazidime ARB, and meropenem residues^a

Parameter	Analysis method	Correlation coefficient
		blaNDM (GCN/ml)	blaKPC (GCN/ml)	blaCTX-M (GCN/ml)	blaSHV (GCN/ml)
intI1 (GCN/ml)	Pearson	−0.063	0.323	0.478	0.795
	Spearman	0.167	−0.071	0.667	0.762
MEM (CFU/ml)	Pearson	0.273	0.984	0.177	0.057
	Spearman	0.024	0.262	0.548	0.571
CAZ (CFU/ml)	Pearson	0.159	0.709	0.551	0.444
	Spearman	0.286	0.238	0.690	0.619
Meropenem (ppb)	Pearson	−0.219	−0.991	−0.050	0.250
	Spearman	−0.500	−1.000	−0.500	0.500

^aThe Pearson and Spearman correlation coefficients which were statistically significant at a P value of <0.05 are bold and underlined in the table. GCN, gene copy numbers; MEM, meropenem; CAZ, ceftazidime.

TABLE 3.

Correlation between sul1, sul2, and dfrA genes and the combination of trimethoprim-and-sulfamethoxazole-resistant ARB and trimethoprim and sulfamethoxazole residues^a

Parameter	Analysis method	Correlation coefficient
		sul1 (GCN/ml)	sul2 (GCN/ml)	dfrA (GCN/ml)
intI1 (GCN/ml)	Pearson	0.640	0.436	0.574
	Spearman	0.929	0.571	0.905
SXT (CFU/ml)	Pearson	0.937	0.994	0.910
	Spearman	0.786	0.833	0.857
Trimethoprim (ppb)	Pearson	−0.126	0.079	−0.136
	Spearman	0.119	0.571	0.262
Sulfamethoxazole (ppb)	Pearson	0.293	0.395	0.189
	Spearman	0.262	0.476	0.405

^aThe Pearson and Spearman correlation coefficients which were statistically significant at a P value of <0.05 are bold and underlined in the table. GCN, gene copy numbers; SXT, trimethoprim and sulfamethoxazole (co-trimoxazole).

TABLE 4.

Correlation between aac(6′)-Ib, qnrA, and qnrB genes, amikacin and ciprofloxacin ARB, and ciprofloxacin residues^a

Parameter	Analysis method	Correlation coefficient
		aac(6′)-Ib (GCN/ml)	qnrA (GCN/ml)	qnrB (GCN/ml)
intI1 (GCN/ml)	Pearson	0.678	0.459	−0.112
	Spearman	0.881	1.000	0.143
AMK (CFU/ml)	Pearson	0.872	0.984	0.165
	Spearman	0.714	0.690	0.595
CIP (CFU/ml)	Pearson	0.903	0.977	0.093
	Spearman	0.810	0.810	0.548
Ciprofloxacin (ppb)	Pearson	−0.279	0.074	0.994
	Spearman	−0.314	−0.314	−0.200

^aThe Pearson and Spearman correlation coefficients which were statistically significant at a P value of <0.05 are bold and underlined in the table. GCN, gene copy numbers; AMK, amikacin; CIP, ciprofloxacin.

DISCUSSION

Hospital wastewater, which receives high loads of antimicrobial agents and human pathogens, is considered a reservoir for antibiotic resistance and other genetic factors which promote the potential spread of AMR to the environment (30, 31). In attempt to control this critical point, this surveillance study aimed at obtaining an in-depth characterization of AMR in hospital wastewater in Singapore by quantifying the presence of antibiotic resistance factors, including antibiotic residues, resistant bacteria, and genetic determinants (i.e., ARGs, integron) in hospital discharge. Local levels of ciprofloxacin, azithromycin, clarithromycin, sulfamethoxazole, and trimethoprim were much lower (∼10-fold) than those reported in another study (12). Sulfamethoxazole and trimethoprim are used in combination as a therapeutic agent (known as co-trimoxazole), and the trend of high concentrations detected at all locations is consistent with most other studies of hospital wastewaters (12). Although hospitals are known to be a source of carbapenems, the levels of meropenem measured in our study were low. The macrolide clarithromycin was detected at higher levels in both H1 samples, which suggests that most inpatients in these two wards had respiratory or skin infections, which are commonly treated with this antibiotic (32). In Singapore, clarithromycin is the standard therapy of community-acquired pneumonia, which is a common diagnosis.

Screening for phenotypic resistance showed that concentrations of ARB differed between locations and ranged between 10⁵ and 10⁶ CFU/ml, which fell within the range determined in another study of hospital wastewater in Iran (33). In our study, we found that average concentrations of ARB across all hospitals and wards indicated that resistance to amikacin, clindamycin, erythromycin, ciprofloxacin, and tetracycline appears to be more prevalent than resistance to meropenem, ceftazidime, vancomycin, chloramphenicol, and co-trimoxazole. Antibiotic prescription trends in local hospitals from 2005 to 2010 reported that a high average daily dose prescription of ciprofloxacin for inpatient users which may result in higher concentrations of ARB resistant to ciprofloxacin (34). The effects of cross-resistance between amikacin and ciprofloxacin have been observed in other studies (35, 36), which offers an explanation for the ARB concentrations observed in both cases.

We found significantly higher concentrations of ARB in location H1a than in the H1b and H2 locations. This implies that although concentrations of specific antibiotics may be high in wastewaters, this does not necessarily translate to high concentrations of bacteria resistant to the antibiotic. The overall abundance of ARB in hospital wastewaters is likely driven by patient-derived shedding and patient demographics in various wards. Wastewater discharge from clinical isolation wards housing patients recovering from severe infections tends to harbor higher levels of ARB detection.

In this study, the average level of intI1 genes detected was at least 2 orders of magnitude lower than that detected in hospital wastewaters in China, with tet and sul genes detected in a lower range as well (37). Class 1 integrons are responsible for the spread and carriage of antibiotic resistance gene cassettes in Gram-negative pathogens (38, 39). The high concentrations of class 1 integrons detected in raw wastewater and the occurrence in ARB isolates were consistent with recent literature reports, e.g., 97% for E. coli isolates in Iran (40), 76% for E. coli and 43% for Salmonella isolates in Netherlands (41), 67% for E. coli isolates in Spain (42), and 63% for E. coli isolates in Thailand (43). This suggests that hospital wastewater is a potential source of class 1 integrons associated with antibiotic resistance. Our results also indicated that the tetracycline [i.e., tet(M) and tet(O)]- and MLS_B [i.e., erm(B) and mef(A/E)]-related ARGs did not display much variation across all sampling locations. The low abundances of qnrA and qnrB in raw wastewater measured by qPCR analysis and in resistance isolates detected by conventional PCR in this study were consistent with other literature reports (44, 45). A survey of hospital wastewaters in Spain indicated levels of bla_TEM, erm(B), qnrS, sul1, and tet(W) of between 10⁵ and 10⁶ gene copies/ml, which is slightly lower than the measured concentrations of ARGs in our study (12). In comparison to the domestic wastewaters in China, the levels of the 4 β-lactamase ARGs in our study (i.e., bla_NDM, bla_KPC, bla_SHV, and bla_CTX-M) were 1 order of magnitude higher than those of the carbapenemase ARGs (i.e., bla_NDM, bla_KPC, bla_OXA, and bla_IMP) detected in China (46, 47). However, the concentrations of total 16S rRNA genes in our hospital wastewaters were slightly lower than those in the domestic wastewaters, which suggests a higher prevalence of these ARGs in hospital wastewaters.

Further investigation of bacterial isolates exhibiting phenotypic resistance to meropenem and ceftazidime confirmed that isolates that were resistant to meropenem were likely carrying the bla_NDM and bla_KPC genes whereas ceftazidime-resistant bacteria were found to be more closely associated with bla_CTX-M and bla_SHV. The bla_NDM gene was more frequently detected in meropenem-resistant bacteria isolated from H1b, including bacterial species such as Acinetobacter junii, Comamonas testosterone, Enterobacter sp., and Pseudomonas sp. In contrast, bla_KPC genes were found to be more prevalent in meropenem-resistant isolates from H1a, which consisted of bacteria belonging to Enterobacteriaceae. This result is consistent with other reports where metallo-β-lactamases was found across various bacterial species and the presence of serine-based carbapenemase was mainly restricted to Enterobacteriaceae (48, 49). The detection of the bla_NDM gene in a variety of bacterial species in hospital wastewaters demonstrates the ability of the bla_NDM-bearing plasmid to successfully establish itself in other strains, providing other possible routes of transmission and dissemination to other bacteria within the hospital wastewater community.

The observed correlations among the detected levels of AB residues, ARB, and ARGs in different sampled locations suggest that some targeted ARGs such as bla_NDM, bla_KPC, sul1, aac(6′)-Ib, and intI1 are well correlated with the resistance mechanism of the relevant antibiotics and that the H1a clinical isolation ward on average had higher concentrations of ARB, ARGs, and intI1 genes. On the basis of the results determined for the cultured ARB, the clinical isolation ward was less diverse and more enriched in members of the Enterobacteriaceae group, which could explain the high concentrations of bla_KPC genes detected in H1a wastewaters and the prevalence of the bla_KPC gene in meropenem-resistant isolates. Bacteria in this group are known to be the main culprits responsible for hospital-associated infections and for the spread of the bla_KPC gene in tertiary care hospitals (50). This work provided an in-depth study of hospital wastewaters as a source of antibiotic resistance in a tropical country by applying both quantitative and qualitative methods to determine resistance factors (i.e., AB, ARB, and ARGs). Although ARB were detected in the wastewater effluents of both hospitals of different ward types, our results indicate that effluents from clinical isolation wards consistently displayed higher loads of ARB and could play a major role in the emergence and spread of antibiotic resistance if not properly treated in the downstream environment. We detected the 16 relevant ARG targets in all hospital wastewater samples where sul and aac(6′)-Ib were among the most abundant genes and cfr was the least. We detected the bla_KPC carbapenem resistance gene in Enterobacteriaceae isolates, while the bla_NDM gene was detected in various species of Enterobacteriaceae, Pseudomonas, and Acinetobacter. This study showed the widespread occurrence of ARB, ARGs, and genetic elements which potentially aid the transfer of antibiotic resistance gene cassette arrays in hospital wastewaters. This underscores the need to ensure that hospital wastewaters are sufficiently treated in wastewater treatment plants to reduce the risk of spread of antimicrobial resistance vectors.

Funding Statement

The funders had no contributions in study design, data collection and interpretation, or the decision to submit the work for publication.